73ef94d 51556a5 a2c94d3 51556a5 a2c94d3
1
2
3
4
5
Reparameterized Absorbing Discrete Diffusion (RADD) small model with t-dce loss trained for 400k iterations. Code: https://github.com/ML-GSAI/RADD. Paper: https://arxiv.org/abs/2406.03736.